I have an audio amp project which, according to simulation, will draw 5.46A peak or 3.56A RMS. Should I dimension the power supply for peak power draw or for RMS?

 

Peak.

 

A large filter capacitor would allow the maximum power supply current to be the average (not RMS) of the amp supply current.

 

It depends on the application.
Firstly, what type of power supply. If it is a switched mode, then it has to supply peak current, so you have to design it for peak current.
If you have a 50/60Hz transformer, then the limit is thermal, and it will then depend on its purpose.
If you are going to play highly compressed rhythm and bass through it, then you will have to design for peak. If you are playing classical which has a high crest factor, probably a quarter of peak will be fine.

 

It never hurts to have more Current-Capacity than is required,
however, depending upon the Amplifiers expected-Load, and the ratings of the Output-Transistors,
You may want to be very careful with the selection of the Power-Supply-Voltage.

Generally, the lower the Impedance that is anticipated, ( ~4-Ohms vs ~8-Ohms ),
the lower the maximum Power-Supply-Voltage is allowed to be
to avoid smoking the Output-Transistors, or the Power-Supply-Transformer.

How much Voltage is the Amplifier designed to handle with the expected-Load ?
.
.
.

 

A large filter capacitor would allow the maximum power supply current to be the average (not RMS) of the amp supply current.

Large, as in, how large?

 

Are going to design this power supply or purchase?

 

Large, as in, how large?

.
How much space do You have ?
Multiple "medium-sized" Capacitors are almost always better than one Big-one.

The Power-Supply can be built into it's own separate Box.
.
.
.

 

Multiple "medium-sized" Capacitors are almost always better than one Big-one.

Why?

 

Are going to design this power supply or purchase?

Could be either way.

 

Adding more capacitance is not always the panacea people might think it is.
Whilst the average current remains the same, twice as much capacitance means twice the pulse current in the rectifier and half the pulse length. That doubles the RMS current. The transformer heating is proportional to the RMS current, so more capacitance = hotter transformer. It also means more pulse currents in the wiring which can find their way into the audio, and therefore more hum.

 

My approach, at least for choosing the transformer, minimum of 50% plus of what I expect the absolute maximum operating power would be. You can go as high as you want with the overdimensioing, if the space and the price is not a problem. Of course I am not suggesting to mount a 10kV transformer.

 

Why?

.
Capacitors have internal-Resistance, usually referred to as "Effective-Series-Resistance" or ESR.
Smaller Capacitors usually have about the same, or slightly less, ESR as a large Capacitor,
so several smaller-Capacitors in parallel will have less overall ESR than one large Capacitor,
which is usually a bonus.
Doing this is standard-practice in Computer-Motherboards,
but is not found as often in Audio-Amplifiers,
usually because of assembly-cost considerations.

Several smaller Capacitors will usually take-up less space than one large Capacitor.

The overall Capacitance, when added together, can remain roughly the same as
the minimum-calculations that are quite often used,
but the total-Capacitance can also be more easily increased.

Adding Inductors before the Capacitors can improve the Power-Supply's performance substantially,
but they can be quite expensive and quite bulky,
they would be roughly the same size as the Main-Transformer, and You would need 2 of them.
2 Inductors will help to stabilize the Power-Supply-Voltage,
and make the Power-Supply run more efficiently, and therefore cooler.
.
.
.

 

REally, it depends on the amplifier power rating standard used. You can save a whole lot by using the IHF Music power rating. That cuts the supply requirements in half.